Sign in Subscribe
Machine learning

Advancements in Robotics: A Probabilistic Formulation of Lidar Mapping with Neural Radiance Fields

Strategic Insights

— 2 min read

Researchers from Tufts University have made a significant breakthrough in robotics and automation by developing a novel method for training Neural Radiance Fields (NeRF) to produce novel LiDAR views of a scene. This innovation, which has been peer-reviewed, allows for the classification of multiple LiDAR returns, reducing the likelihood of "phantom surfaces" in free space. By formulating loss as an integral of probability, the network can learn multiple peaks for a given ray, enabling the sampling of first, nth, or strongest returns from a single output channel.

Key Takeaways:

  • The researchers reexamined the process of training NeRF to produce novel LiDAR views of a scene, recognizing the challenges posed by the probabilistic nature of LiDAR returns.
  • The team formulated loss as an integral of probability, allowing the network to learn multiple peaks for a given ray and reducing the occurrence of "phantom surfaces" in free space.
  • The probabilistic formulation enables the sampling of first, nth, or strongest returns from a single output channel, improving the accuracy and efficiency of LiDAR mapping.
  • The research was supported by the U.S. Department of Transportation Joint Program Office (ITS JPO) and the Office of the Assistant Secretary for Research and Technology (OST-R).
  • The study has been published in IEEE Robotics and Automation Letters and has been recognized as a key innovation in the field of robotics and automation.
  • The researchers, led by Matthew Mcdermott from Tufts University, Department of Mechanical Engineering, aim to further develop this technology for real-world applications.

Statistics:

  • The research was published in IEEE Robotics and Automation Letters in 2025, with a volume of 10(6) and page numbers 5409-5416.
  • The study concluded that the probabilistic formulation yields improved results in LiDAR mapping, with an increase in accuracy and efficiency.
  • The research received support from the U.S. Department of Transportation Joint Program Office (ITS JPO) and the Office of the Assistant Secretary for Research and Technology (OST-R).
  • The team of researchers, led by Matthew Mcdermott, has published several papers on related topics, showcasing their expertise in robotics and automation.

Sources:

  • A Probabilistic Formulation of Lidar Mapping With Neural Radiance Fields, published in IEEE Robotics and Automation Letters (Ieee-inst Electrical Electronics Engineers Inc, 2025)
  • Ieee Robotics and Automation Letters, Volume 10, Issue 6, 2025, pages 5409-5416
  • IEEE-inst Electrical Electronics Engineers Inc, 445 Hoes Lane, Piscataway, NJ 08855-4141, USA
  • NewsRx. New Robotics and Automation Findings from Tufts University Reported (A Probabilistic Formulation of Lidar Mapping With Neural Radiance Fields). Robotics & Machine Learning. June 23, 2025; p 228.

Read more